Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use a five-network deep ensemble instead of MC-dropout | probabilistic neural networks · Boston housing | NLL: 2.46 $\pm$ 0.25 -> 2.41 $\pm$ 0.25 | Helped |
| Use a five-network deep ensemble instead of MC-dropout | probabilistic neural networks · Concrete | NLL: 3.04 $\pm$ 0.09 -> 3.06 $\pm$ 0.18 | Hurt |
| Use a five-network deep ensemble instead of MC-dropout | probabilistic neural networks · Energy | NLL: 1.99 $\pm$ 0.09 -> 1.38 $\pm$ 0.22 | Helped |
| Use a five-network deep ensemble instead of MC-dropout | probabilistic neural networks · Kin8nm | NLL: -0.95 $\pm$ 0.03 -> -1.20 $\pm$ 0.02 | Helped |
| Use a five-network deep ensemble instead of MC-dropout | probabilistic neural networks · Naval propulsion plant | NLL: -3.80 $\pm$ 0.05 -> -5.63 $\pm$ 0.05 | Helped |
| Use a five-network deep ensemble instead of MC-dropout | probabilistic neural networks · Power plant | NLL: 2.80 $\pm$ 0.05 -> 2.79 $\pm$ 0.04 | Helped |
| Use a five-network deep ensemble instead of MC-dropout | probabilistic neural networks · Protein | NLL: 2.89 $\pm$ 0.01 -> 2.83 $\pm$ 0.02 | Helped |
| Use a five-network deep ensemble instead of MC-dropout | probabilistic neural networks · Wine | NLL: 0.93 $\pm$ 0.06 -> 0.94 $\pm$ 0.12 | Hurt |
| Use a five-network deep ensemble instead of MC-dropout | probabilistic neural networks · Yacht | NLL: 1.55 $\pm$ 0.12 -> 1.18 $\pm$ 0.21 | Helped |
| Use a five-network deep ensemble instead of MC-dropout | probabilistic neural networks · Year Prediction MSD | NLL: 3.59 $\pm$ NA -> 3.35 $\pm$ NA | Helped |
| Add an ensemble of five networks that learn predictive variance (ablation) | single-network maximum-likelihood regression · Boston housing | NLL: 2.55 $\pm$ 0.36 -> 2.41 $\pm$ 0.25 | Helped |
| Add adversarial training (ablation) | single-network maximum-likelihood regression · Naval propulsion plant | NLL: -5.65 $\pm$ 0.28 -> -4.08 $\pm$ 0.13 | Hurt |
Closest papers by meaning
- Conservative Uncertainty Estimation By Fitting Prior Networks (ICLR 2020 · similarity 0.91)
- Depth Uncertainty in Neural Networks (NeurIPS 2020 · similarity 0.91)
- Make Me a BNN: A Simple Strategy for Estimating Bayesian Uncertainty from Pre-trained Models (CVPR 2024 · similarity 0.90)
- SURE: SUrvey REcipes for building reliable and robust deep networks (CVPR 2024 · similarity 0.90)
- Uncertainty Quantification for Deep Regression using Contextualised Normalizing Flows (NeurIPS 2025 · similarity 0.90)
- Uncertainty Quantification with the Empirical Neural Tangent Kernel (NeurIPS 2025 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.